We will be hosting the 4^th International PyomoFest in Austin, TX on January 14, 2017 – the day immediately preceding the INFORMS Computing Society (ICS) meeting. 

The purpose of the PyomoFest workshop series is three-fold:

1)       To provide an opportunity for Pyomo developers and end-users to share their collective experiences.

2)       To update the Pyomo user community on the latest and upcoming developers in Pyomo.

3)       To gather feedback from the end-user community regarding future Pyomo development.

This workshop follows the inaugural workshop in Santa Fe (in 2015) and PyomoFest Santiago (Chile) and Aachen (Germany).

The workshop will be organized as follows. The morning will be dedicated to a brief overview / tutorial for new Pyomo users, followed by a series of presentations by core Pyomo developers describing new Pyomo features (e.g., pyomo.dae) and previews of upcoming features (e.g., unit specification). The afternoon will consist of a series of “contributed” end-user talks, falling into one of the following categories:

·         Application specific case studies based on Pyomo.

·         End-user-developed extensions of Pyomo.

·         Third-party applications libraries based on Pyomo.

·         Experiences with Pyomo in the classroom.

·         Rants / philosophical musing about Pyomo and algebraic modeling languages in general.

Plenary Talk by Prof. Martin Wainwright.

Large-scale data sets are now ubiquitous throughout engineering and science, and present a number of interesting challenges at the interface between statistics and optimization.  This talk provides two vignettes in this area.  The first concerns the use of randomized dimensionality reduction techniques, also known as sketching, for obtaining fast but approximate solutions to large-scale convex programs.  We describe a version of sketching that adapts to the intrinsic dimension of the solution space, and show how it leads to randomized version of the Newton algorithm with strong guarantees. The second vignette focuses on low-rank matrix estimation using nonconvex programs, for which it is nonetheless possible to obtain rigorous guarantees on the solution quality.

Tutorial by Prof. Shane Henderson

Cornell’s School of Operations Research and Information Engineering has been working with the bike-sharing company Citibike since Citibike began operations in New York City in 2013. We provide data analysis and advice about strategy and operations, not just to Citibike, but also to its parent company Motivate that operates several bike-sharing programs around the USA. I’ll describe some of our modeling work with Citibike, focusing on a suite of models (not just simulation models) that informs the decision about where to position both racks and bikes around the approximately 500 stations in NYC. Our white-box simulation-optimization algorithms can be viewed as heuristics, but they push the state of the art in simulation optimization, and the central principle applies to many other settings.

Plenary Talk by Prof. Dimitri Bertsekas

Abstract: We will review connections between two currently very active research fields, focusing primarily on large linear fixed point problems. We show that, under certain assumptions, there is a close relation between two types of methods: 1) Proximal algorithms, which are prominent in numerical convex analysis and optimization, and 2) Multistep methods of the temporal difference type such as TD(lambda), LSTD(lambda), and LSPE(lambda), which are central in simulation-based approximate dynamic programming, notable for its recent successes with programs that play games, such as Go, Backgammon, and others, at impressive and sometimes above human level. As an application of this connection,  we provide a new and simple way to accelerate the standard proximal algorithm by extrapolation towards the multistep iteration, which generically has a faster convergence rate. Moreover, we use the connection with multistep methods to integrate into the proximal algorithmic context several new ideas that have emerged in the approximate dynamic programming context. In particular, we consider algorithms that  project each proximal iterate onto the subspace spanned by a small number of  basis functions, using low-dimensional calculations and simulation, and we discuss various algorithmic options from approximate dynamic programming.

Tutorial by Prof. Cole Smith

Research interest and publications regarding network interdiction problems have expanded at a dizzying pace in recent years. This tutorial is designed to briefly introduce the field of interdiction problems and to illustrate some areas of future research based recent developments appearing in the literature and at this conference. We first cover classical algorithms, particularly those that rely on ad-hoc formulations, duality-based reformulations, and cutting-plane algorithms. Next, we focus on emerging algorithms, variations on interdiction problems, and new applications for interdiction. The algorithms we will discuss include new sampling techniques for interdiction and fortification, along with approaches that exploit special structures in interdiction problems. Some recent variations on interdiction problems include diversion problems where routes are forced onto a particular arc, and dynamic interdiction problems in which the attacker and defender repeat their moves sequentially. Finally, the tutorial will investigate new areas in which interdiction applications are being used, including logistics, cybersecurity, and critical infrastructure protection.

Tutorial by Prof. Andrew Schaefer